Cost-Efficient Task Scheduling in Cloud Computing Environments Using Machine Learning Techniques
|has title::Cost-Efficient Task Scheduling in Cloud Computing Environments Using Machine Learning Techniques|
|Master:||project within::High Performance Distributed Computing|
|Student name:||student name::Stefan Bantea|
|Second reader:||has second reader::Guszti Eiben|
As cloud computing develops, application developers are presented with options regarding the performance capabilities and costs of the available machines. Since statically choosing one type of machine typically proves to lead to inefficiency, performance-wise or cost-wise, dynamically deciding at run-time which machines to use promises to improve this aspect.
The goal of this project is to create a cost-efficient schedule for cloud computing environments by using machine learning techniques such as locally weighted learning, neural networks and evolutionary algorithms. The basic idea is to predict the execution times of jobs on each type of machine and then create schedules that optimize the overall cost and run time.
The applications used for testing the scheduler will satisfy some basic requirements: over 50000 jobs, complex execution time distribution, at least 5 command-line parameters that have an impact on the execution time of the jobs. Usage of benchmark applications is strongly encouraged.